Dendritic Growth of Two-Dimensional Protein Crystals - American

Alice P. Gat*,+. Department of Chemical Engineering, Stanford University, Stanford, California 94305, and. Beckman Laboratories for Structural Biology...
0 downloads 0 Views 4MB Size
Langmuir 1992,8,2357-2360

2357

Dendritic Growth of Two-Dimensional Protein Crystals Andrew C. Ku; Seth A. Darst,t Roger D. Kornberg,t Channing R. Robertson3 and Alice P. Gat*,+ Department of Chemical Engineering, Stanford University, Stanford, California 94305, and Beckman Laboratories for Structural Biology, Department of Cell Biology, Fairchild Center, Stanford University, Stanford, California 94305 Received April 30,1992. In Final Form: July 16,1992 Two-dimensionalcrystals of the protein streptavidinform on lipid monolayers containingbiotinylated lipids. Introducingavidin as a noncrystallizablecontaminant induced dendriticgrowth of the streptavidin crystals. Increasing the proportion of avidin to streptavidin shifted the dendritic streptavidin crystal morphology from amorphous fingers to needlelike structures. These results provide insight into the mechanism and dynamics of the two-dimensionalprotein crystallization process. Dendritic crystal growth is an example of spontaneous pattern generation in nonequilibrium systems.' While three-dimensionaldendrites have been grown and analyzed under controlled conditions: efforts have also been made to simplify the analysis by reduction of dimensionality, through growth in a narrow gap between parallel plates.3 The results were only partially in accord with theoretical expectations,due in part to the dendrites remainingthreedimensionaldespite the restricted geometry, and wetting of the plate^.^ We report here on dendritic crystal growth system, with the further in a quasi-two-dimensional(2-D) advantage that crystal growth can be monitored at the molecularlevel, and the intermolecular contacts are known. This work complements research in phase transitions of lipid and fatty acid monolayers where condensed-phase domains have been observed to assume a rich variety of shapes including circular, spiral, and unstable branched structure^.^ Our approach should afford insight into the mechanism of dendrite formation as well as other aspects of crystal growth involving protein-lipid associated systems. True 2-D crystals of proteins can be formed on lipid layers containing ligands for protein binding.6 Crystallization is facilitated by lateral diffusion of the lipids and by orientation due to the specificity of protein binding. The example best suited to our present purposes is the crystallization of the bacterial protein streptavidin on layers of biotinylated lipid^.^-^ Large 2-D crystals form readily, and both the arrangement of protein in the crystals and full three-dimensional structure of the protein have been determined. Crystals formed from fluorescently labeled streptavidin are visible in the light microscope; they are typically hundreds of micrometers in extent, and

* To whom correspondenceshould be addressed. t Department t Department

of Chemical Engineering. of Cell Biology. (1)Kessler, D. A.; Koplik, J.; Levine, H. Phys. Rev. A 1986,33,3352. (2)Glicksman, M. E.;Shaefer, R. J.; Ayers, J. D. Metall. Trans. A 1976,7,1747.Huang, S.-C.;Glicksman, M. E. ActaMetall. 1981,29,701. Somboonsuk, K.; Mason, J. T.;Trivedi, R. Metall. Trans A 1984,15,967. (3)Honjo, H.; Ohta, S.; Sawada, Y. Phys. Rev. Lett. 1985,55, 841. Dougherty, A.; Gollub, J. P. Phys. Rev. A 1988,38, 3043. Chou, H.; Cummins, H. Z. Phys. Rev. Lett. 1988,61,173. Maurer, J.; Bouissou, P.; Perrin, B.; Tabeling, P. Europhys. Lett. 1989,8,67. (4)Fattinger, Ch.;Honegger, F.; Lukosz, W. Phys. Rev. Lett. 1986,57, 2536. (5)Gaub, H. E.;Moy, V. T.; McConnell H. M. J. Phys. Chem. 1986, 90,1721. Miller, A.; Knoll, W.; Mijhwald, k.Phys. Rev. Lett. 1986,56, 2633. Suresh, K. A.; Nittmann, J.; Rondelez, F. Rrogr. Colloid Polym. Sci. 1989,79, 184. (6)Komberg, R. D.; Darst, S. A. Curr.Opin. Struct. Biol. 1991,I,642. (7)Blankenburg,R.; Meller, P.; Ringsdorf,H.; Salesse, C. Biochemistry 1989,28,8214. (8)Darst, S. A.; et al. Biophys. J. 1991,59,387. (9)Kubalek, E. W.;Komberg, R. D.; Darst, S. A. Ultramicroscopy 1991,35,295.

0743-7463/92/2408-2357$03.00/0

A

B

C

Figure 1. SDS polyacrylamide gel electrophoresis (15%acryImide, Bio-Rad) of three commercial preparations of streptavidin. (A) This preparation (Sigma, lot no. 110H68351),comprised of a single polypeptide, formed abundant 2-D crystals, largely nondendritic, covering over 90% of the lipid surface. (B) This preparation (Sigma, lot no. 119F-6806) included two polypeptides of larger apparent molecular weight and formed fewer crystals, all highly dendritic. (C)The final preparation (Sigma, lot no. 80H6811) barely crystallized at all, and was comprised of the two larger polypeptides present in (B) as well as a third peptide, larger still.

frequently exhibit rectangular, H-like shapes. We began with the notion that the lobes of the H-shaped crystals were precursors to dendrites, initiated by a crystal growth instability, and here we present results supporting this hypothesis. An experimental approach was suggested by observations with varouscommercialpreparations of streptavidin. One preparation formed abundant 2-D crystals, largely nondendritic and covering over 90% of the lipid surface, a second preparation formed fewer crystals, all highly dendritic, and a third preparation barely crystallized at all. SDS gel electrophoresis (Figure 1)revealed a single polypeptide in the f i t preparation, an additional polypep tide of larger apparent molecular weight in the second preparation, and another larger polypeptide in the third preparation. It was previously reportedlothat proteolysis of streptavidin at the N-and C-termini, from the native 159aminoacid polypeptide to a "core" of 121-127 residues, facilitates the growth of three-dimensional crystals. We (10)Pahler, A.; Hendrickson, W. A.; Kilks, M. A. G.;Argarana, C.E.; Cantor, C. R. J. Biol. Chem. 1987,262,13933.Weber, P. C.; Ohlendorf, D. H.; Wendoloski, J. J.; Salemme, F. R. Science 1989,243,85.

1992 American Chemical Society

Letters

2358 Langmuir, VoZ. 8,No. 10,1992

Figure 2. Fluorescence micrographs of 2-D crystals of FITC-streptavidin grown at the air-water interface on a lipid monolayer of 2.3 mol % N- [6-(biotinoylamino)hexanoyl]dipalmitoyl-~-a-phosphatidylethanolamine (Molecular Probes) in dioleoylphosphatidylethanolamine (Avanti Polar Lipids) at 25-30 dyn/cm. Lipid solution at 1mg/mL (in chloroform/hexane, 1:l vol/vol) containing 2.3 mol % N- [6-(biotinoylamino)hexanoyl]dipalmitoyl-L-a-phosphatidylethanolaminein dioleoylphosphatidylethanolamine was spread at the air-water interface of a miniature Langmuir trough. Streptavidin (Boehringer Mannheim, homogeneous and similar in molecular weight to preparation A described in Figure 1)and avidin (Sigma) were labeled with approximately two FITC and TRITC per protein, respectively.28Crystallizationwas performed at a lipid surface pressure of 25-30 dyn/cm and room temperature (21 f 1 "C). Crystallinity was confirmed with two tests. Samples transferred onto electron microscope grids and then stained with uranyl acetate revealed molecules arranged in well-ordered arrays, and under linearly polarized excitation light, the fluorescent domains exhibited optical anisotropy. Scale bars represent 100 pm. The fluorescence micrograph on the left (FITC excitation) is of a 2-D FITC-streptavidin crystal, 20 mol % FITC-streptavidin, 80 mol % TRITC-avidin after 61 min. The arrows indicate regions of interaction between dendrites, curtailing growth. The image was recorded using a linear polarizer in the excitation path. The fluorescence micrograph on the right (TRITC excitation) is of the same 2-D FITC-streptavidin crystals as those on the left, surrounded by TRITC-avidin.

surmisedthat the polypeptide in the first preparation had undergone sufficient proteolysisto crystallizewell in 2-D, whilethe larger polypeptide in the other preparations failed to crystallize. The formation of dendrites occurred in a mixture of the larger protein with the proteolyzed form as in the second preparation or when the first and third samples were mixed. The larger polypeptide represents a noncrystallizing impurity, and in a transport-limited system, such impurities diffuse too slowly from the advancing crystal front to prevent their accumulation at the front. A protrusion on the surface of the crystal enters a region of lower impurity concentration and grows more rapidly than the rest of the front. Similar diffusioncontrolled Mullins-Sekerkall growth instabilities have been observedfor pressure-or temperature-induced phase transitions in planar monolayers of lipid/dye12and fatty acid/dye.l3 Modeling of this diffusion-limitedaggregation led both groups to concludethat the branched, self-similar patterns were due to constitutional undercooling. To test the idea that a noncrystallizing streptavidin polypeptide was responsible for dendrite formation, and to develop a system for controlled dendritic growth, we investigated the crystallizationof core streptavidin labeled with fluorescein isothiocyanate (FITC; Sigma) in the presence of unlabeled hen egg-white avidin. Although avidin is similar in size and affinity for biotin to streptavidin,14it bears surface polysaccharidesknown to hinder (11) Mullins, W. W.; Sekerka, R. F. J. Appl. Phys. 1963,34, 323. (12) Miller, A.; Mdhwald, H. J . Chem. Phys. 1987,86, 4258. (13) Suresh, K.A.;Nittmann, J.; Rondelez, F. Europhys. Lett. 1988, 6, 437. (14) Green, N.M. Adu. Protein Chem. 1975, 29, 85.

~rystal1ization.l~Indeed, no crystallization of avidin labeled with tetramethylrhodamine isothiocyanate (TRITC; Molecular Probes) could be detected by light microscopy, and only small crystallites were observed in the electronmicroscope. Avidinwould therefore constitute a noncrystallizingimpurity in a mixture with.,corestreptavidin. To confirm that avidin was preferentiallyexcluded from the ordered phase, we examined the growth of FITCstreptavidin crystals mixed with TRITC-labeled avidin. When observing the tetramethylrhodamine fluorescence, the FITC-streptavidin crystals appeared less fluorescent than the surrounding, disordered region, as illustrated in Figure 2, indicating that the TRITC-avidin was preferentiallyexcludedfrom the streptavidin crystals. To ensure that crystallizationwas not altered by the fluorescentlabel, we performed experiments with the reverse labeling scheme. The same crystal morphologies were observed. We altered the macroscopic features of the 2-D streptavidin crystalsby mixing avidin in varying proportionswith streptavidin at fixed total protein comentration. Dendrites grown at 15,25, and 40 mol % streptavidin varied in morphology from needle-likeat low streptavidin (Figure 3, left, middle) concentrations to less highly ramified at higher concentrations (Figure 3, right). These structures illustrate the effects of crystal growth instabilities arising from transport-limited crystallization.16 Sincethe crystal growth rate depends on both the mole percent streptavidin in the system and the crystallization time, we selected micrographs in Figure 3 to show representative crystals (15) Gatti, G.; et al. J. Mol. B i d . 1989, 78,787. Bruch, R.C.;White,

H.B. Biochemistry 1982,21, 5334.

(16) Langer, J. S. Reu. Mod. Phys. 1980,52, 1.

Letters

Langmuir, Vol. 8, No. 10, 1992 2359

Figure3. Fluorescencemicrographs(FITCexcitation)illustratingmorphologiesof 2-Dstreptavidincrystalsat three (bulk)concentrations (scale bars 100 pm). (Left) This micrograph is of 15 mol % FITC-streptavidin, 85 mol % TRITC-avidin; t = 49 min. Crystals are needlelike with limited sidebranching. (Middle) The micrograph is of 25 mol % FITC-streptavidin,75 mol % TRITC-avidin; t = 36 min. Crystals have more extensive sidebranching. (Right) This micrograph is of 40 mol % FITC-streptavidin, 60 mol % TRITCavidin; t = 21 min. Crystals are broad and amorphous. Arrows indicate exceptionally large sidebranches adjacent to smaller ones.

at times just beyond the reduction of tip velocity, ranging from 8 min at 40 mol % streptavidin to 55 min at 10 mol % streptavidin. Crystal morphologies vary in a regular fashionwith the concentrationof protein. Increasingmole percent streptavidin concentration yields crystals growing large enough to observe the instability through several generations. We note the appearance of secondary dendrites and more extensive sidebranching as we increase the mole percent concentration; ternary branches appear in crystals grown at 40% streptavidin (Figure 3, right). The dendrites initially grew at a constant velocity. We monitored their growth photographicallyand determined primary dendrite tip speeds. The tip speed increased linearlywith mole percent streptavidin concentration; tip speeds ranged from 1.4 f 0.3 pm/min at 10 mol % streptavidin to 50 f 8 pmlmin at 40 mol % streptavidin (data not shown). Extrapolating to zero velocity yields a crystallization threshold of 9 mol % streptavidin. At lower streptavidin concentrations, where there was sufficient area for the dendrites to grow spatially unhindered, the dendrites protruded at characteristic angles between 35O and 50' (Figure 3, left). Typically, once ah angle between primary dendrites was selected for a particular crystal, it was preserved as secondary dendrites developed. From electron microscopyof long crystallites, two perpendicular growth directions appear, one of which is preferred (unpublished results). Because the macroscopic angle is not 90° but instead ranges from 35O to 50°, we hypothesize that dendrites grow dong a linear combination of these two growth directions. In contrast to the well-preservedangles observedat low streptavidin mole percent concentrations, higher mole percent streptavidin concentrations led to more amorphous primary and secondarydendrites,failingto sustain a characteristicangle (Figure3, right). This is attributable in part to the spatial constraints imposed by crowding,causing the brmches to bend or deform. Our protein crystal morphologies closely resemble the solid-phasedomains formed by compressing a diacetylenelipid monolayer, as studied by Giibel et a1.T

the anisotropic morphologiespossess four lobes extending at characteristic angles. Furthermore, solid diacetylene lipid domains can be photopolymerized,and the resultant structures exhibit optical anisotropy. Under appropriate conditions, abnormally large sidebrancheswere observed adjacent to short sidebranches as shown by the arrows in Figure 3, right. When crystals grow sufficientlyclose together, their local diffusion fields interfere, leading to competitionfor included protein. This competition can cause adjacent branches to differ greatly in size, reminiscent of structures created by viscous fingering systems.18 The growth rates of dendrite tips approaching other growing crystals were severely diminished (Figure 2, left, arrows), while other dendrite tips on the same crystal continued to grow. The four primary dendrites of each crpstal thus appear to grow independently. A t long times, well after primary dendrites ceased to grow outward, secondarydendrites slowly coarsened.The sidebranches evolved from fingerlike to rectangular protrusions resembling a series of tethered blocks (Figure 4). These rectangles are reminiscent of the rectangular H-like streptavidin crystals observed previ0usly.~*8This phenomenon may be a 2-Danalog of sidebranch coarsening, first observed by Huang and Glicksman for solidifying succinonitrileduring isothermalholding.l9 They proposed that the coarsening is likely driven by interfacial energy reduction. Comparable annealing of branched domains to a final, compact equilibrium shape has also been observed with lipid and fatty acid mon01ayers.~~J3,20 In these 2-D cases, the shape evolution is believed to arise from the minimizationof the total line free energy. Further studies tracking sidebranches over time at higher magnification are in progress. The boundary layer of impurity (i.e., avidin) associated with the advancing crystal front is characterized by the (17) Gobel, H. D.;Gaub, H. E.; Mohwald, H. Chem. Phys. Lett. 1987, 138,441. (18) Homsy, G. M.Annu. Rev. Fluid Mech. 1987,19,271. (19) Huang, S.-C.; Glicksman, M. E. Acta Metall. 1981,29, 717. (20) Akamatsu, S.;Rondelez, F. J . Phys. II 1991,1, 1309.

2360 Langmuir, Vol. 8, No. 10,1992

Letters

I Figure 4. Fluorescence micrographs (TRITC excitation) of a 2-D streptavidin crystal. Micrographs are presented in false color to emphasize detail (scale bars 100 pm). (Left) This micrograph is of 10 mol % FITC-streptavidin, 90 mol % ,TRITC-avidin; t = 52 min. Crystals are needlelike with extensive sidebranching. (Right) This micrograph is of 10 mol % FITC-streptavidin, 90 mol 5% TRITC-avidin; t = 5 h and 23 min. Sidebranches have coarsened.

diffusion length scale I = 26)/v, where 6) is the lateral diffusivity of bound streptavidin and u is the dendrite tip speed. An estimate of the diffusivitycan be approximated from work with other 2-D systems. At dilute concentrations, lateral diffusion coefficients for lipid haptenantibody complexes in a planar model membrane were the same as those of the lipid molecules themselves.21By increasingthe haptenated lipid concentration from 0.5 to 25 mol %, the lateral diffusivity of bound antibody decreased 2 orders of magnitude.22The membrane protein bacteriorhodopsin shows a decrease in lateral diffusivity by a factor of 20 from dilute to high protein concentrati0ns.~3 Monte Carlo simulations demonstrate that an increase in fractional surface coverage from 0% to 70% leads to a 10-20-fold decrease in diffusivity values.% We estimate a 100-fold decrease in diffusivity as an upper bound for our system due to the high concentration in an interface saturated with streptavidin and avidin. The “diluent”lipid in our system,dioleoylphosphatidylcholine, has a lateral diffusivity of 5 X cm2/s at our working surface pressure.25 Consequently, we estimate the diffusivity of the bound protein as 5 X cm2/s. Using this diffusivity,we estimate a lower bound for I of 40 pm at 10 mol % streptavidin to 1pm at 40 mol %. This diffusionlength exceeds the radius of curvature of the dendrite tip by 1order of magnitude. Similar ratios of diffusion lengths to radius of curvature have been observed in three-dimensional dendritic systems.26 We have noted faint halos of decreased fluorescence intensity surrounding quickly growing FITC-streptavidin crystals, suggesting impurity-rich boundary layers; the thickness (21) Smith, L. M.; Parce, J. W.; Smith, B. A.; McConnell, H. M. R o c . Natl. Acad. Sci. U.S.A. 1979, 76, 4177. (22) Subramaniam, S. Ph.D. Thesis, Stanford University, Stanford, CA, 1987. (23) Peters, R.; Cherry, R. J. R o c . Natl. Acad. Sci. U.S.A. 1982, 79, 4317. (24) Pink, D. A. Biochim. Biophys. Acta 1985,818,200. Saxton, M. J. Biophys. J. 1987, 52, 989. Minton, A. P. Biophys. J. 1989, 55, 805. (25) Bohorquez, M.; Patterson, L. K. J. Phys. Chem. 1988,92,1835. (26) Ben-Jacob,E.;Goldenfeld,N.; Langer,J. S.;Schon, G. Phys. Reu. A 1984,29,330.

of these halos compares favorably with the estimated diffusion lengths. Work is underway to quantify these features and to compare this experimental system with predicted 2-D dendritic growth model^.^^^^ Success in comparing model predictions with experimental observations is typified by studies by Miller and Mohwald12 using small quantities of TPyP dye to visualize the twophase liquid-condensedlliquid-expandedregime of the lipid dimyristoylphosphatidylethanolamine.Partitioning of dye (acting as impurity) into the adjacent liquidexpanded phase yielded halos of dye-enriched regions surrounding the liquid-condensed domains. There was qualitative agreement between experimental and calculated dye concentration profiles based on diffusioncontrolled growth. In summary, we have discovered a new, quasi-twodimensional, crystallization instability that has features similar to those of the more often studied three-dimensional growth patterns. The ability to manipulate the morphology of such crystalswith the addition of analogous but uncrystallizableimpurities has allowed us to illustrate the characteristics of the growth process. We observe the creation of three generations of dendrites, the competitive growth in neighboring crystals, and the preservation of growth directionthrough successivedendrites. We believe that now pattern formation and crystallization growth processescan be studied in a systematicway in a simplified system of reduced dimensionality.

Acknowledgment. We thank E. Kubalek for assistance with electron microscopy and J. Pierce for comments on the paper. This work was funded in part by the’Natural Science and Engineering Research Council of Canada. RegistryNo. Streptavidin,9013-20-1;N-(6-(biotinoy1amino)hexanoy1)dipalmitoyl-L-phosphatidylethanolamine, 119830-818; dioleoylphosphatidylethanolamine, 2462-63-7. (27) Meiron, D. I. Phys. Reu. A 1986,33, 2704. Ben Amar, M.; Pelc6, P. Phys. Rev. A 1989,39,4263. Kessler,D. A.; Levine, H. Phys. Rev. Lett. 1991,67,3121. (28) Nargessi, R. D.; Smith, D. S. Methods Enzymol. 1986,122,67.